Skip to main content
Technical deep dives are recorded sessions (~60 minutes) where we explore advanced topics, walk through hands-on implementations, and dive deep into specific use cases with LFM models. These sessions are designed for developers who want to learn advanced techniques and best practices.

Recorded sessions

Watch our previous technical deep dive sessions to learn from real-world implementations:

Fine-tuning LFM2-VL for image classification

November 6, 2025 — Learn how to fine-tune LFM2-VL models for custom image classification tasks, including dataset preparation, training strategies, and deployment.

Building a 100% local audio-to-speech CLI with LFM2-Audio

November 27, 2025 — Build a completely local audio transcription CLI tool using LFM2-Audio models without any cloud dependencies.

Fine-tuning LFM2-350M for browser control with GRPO and OpenEnv

December 26, 2025 — Advanced reinforcement learning session on training browser control agents using Group Relative Policy Optimization.

Local video-captioning with LFM2.5-VL-1.6B and WebGPU

January 22, 2026 — Implement real-time video captioning running entirely in the browser using WebGPU for hardware acceleration.

What to expect

Each technical deep dive session includes:

Live coding

Watch as we build real projects from scratch with step-by-step explanations

Best practices

Learn recommended approaches for fine-tuning, deployment, and optimization

Q&A sessions

Get your questions answered by the Liquid AI team and community experts

Code samples

Access the complete code and resources used in each session

Topics covered

Our technical deep dives cover a wide range of advanced topics:
  • Fine-tuning techniques — Supervised fine-tuning, GRPO, DPO, and LoRA
  • Vision-language models — Image classification, object detection, video understanding
  • Audio models — Transcription, speech recognition, audio processing
  • Deployment strategies — Browser-based deployment, mobile optimization, edge devices
  • Advanced architectures — Mixture of agents, agentic workflows, tool calling
  • Performance optimization — Quantization, caching, hardware acceleration

Join live events

Want to attend the next technical deep dive session live?

Join #live-events on Discord

Head to the #live-events channel on our Discord server to get notified about upcoming sessions, ask questions during live streams, and connect with other developers.
Live sessions give you the opportunity to:
  • Ask questions in real-time
  • Influence the direction of the session
  • Network with other developers
  • Get early access to new features and techniques

Request a topic

Have a topic you’d like us to cover in a future technical deep dive? Let us know!
1

Join Discord

Join our Discord community if you haven’t already
2

Share your idea

Post your topic suggestion in the #live-events channel with details about what you’d like to learn
3

Vote and discuss

Community members can vote on suggested topics and discuss what would be most valuable

Additional resources

Complement your learning from technical deep dives with these resources:

Community projects

Explore real-world projects built by the community

Deployment options

Learn about different ways to deploy LFM models

Fine-tuning guide

Master fine-tuning techniques for custom use cases

Official docs

Read the complete Liquid AI documentation

Build docs developers (and LLMs) love